Application of deep learning in bioassessment of aquatic ecosystems: toward the construction of automatic identifier of aquatic macroinvertebrates
Abstract: Biodiversity loss and degradation of aquatic ecosystems have been accelerating worldwide over the last decades. To halt or even revert this process is of pivotal importance for sustainable development. A cost-effective biomonitoring system is a basis for any conservation management, through which comprehensive taxa lists, ideally at high taxonomic resolution (i.e., species level), are provided. The identification of biotic material as the first step in this process is hindered when the traditional morphology-based approach is applied. This is especially true for „dark taxa“, such as non-biting midges (Chironomidae, Diptera), a group of species which is ecologically abundant and important, yet their identification is difficult, time-consuming, and requires high expertise. Having all this in mind, combining morphology-, DNA- and deep learning-based approaches simultaneously on aquatic macroinvertebrate samples we wanted to build a new application for artificial intelligence-based identification of species. This is one of the first implementations of artificial intelligence and deep learning method in the biomonitoring world. To realize the main objective of the project, two hundred macroinvertebrate species, encompassing chironomids and EPT group, from the South Morava river basin and the Danube, morphologically identified and validated by DNA taxonomy, were used for the construction of the deep learning model. The main project deliverables are web-based and standalone applications suitable to end-users that employ the obtained deep learning model. The results of the project facilitate the identification process of macroinvertebrates, as a key group for aquatic ecosystem monitoring, via machine learning and enable their cost-effective implementation in routine bioassessment programs. This research was conducted under the AIAQUAMI project number 7751676 funded by the Science Fund of the Republic of Serbia through the programme IDEJE_PN; provided from the budget of the Republic of Serbia and the World Bank project - the Serbia Accelerating Innovation and Entrepreneurship Project (SAIGE).
Homepage Prof Djuradj Milošević
IGB Colloquia open up!
IGB strives to facilitate and accelerate the exchange of knowledge and ideas within and also outside of IGB. One element contributing to inter- and transdisciplinary exchange, and more (scientific) cooperation and innovation, is to open up IGB Colloquia to an interested external audience from science (other research institutes, universities, laboratories) as well as practice (i.e. conservationists, freshwater/land-use managers, authorities, associations). If you would like to join this IGB Colloquium as a guest, we ask you to register until the morning of the colloquium 10 o'clock the latest. After we have checked your registration, you will receive the participation link.